On Thursday 26 October 2006 22:55, Joel E. Denny wrote: > On Thu, 26 Oct 2006 [EMAIL PROTECTED] wrote: > > > That is, do you believe this warning is too much trouble to be > > > worthwhile? > > > > The warning would be ok, if it was only true. > > It's true inasmuch as we specifically intended to warn about the current > rule's actions without regard to other rules. Perhaps that would be clear > if the message were reworded. even if you reword it, it will still be missleading.
> > > > Surely it's useful in many cases, but could something be > > > changed about it to accommodate the troublesome cases? Or is it ok as > > > is? > > > > the best thing would to detect that the $3 in rule procoptionlist is > > referenced as $0 in procoption > > In general, that seems difficult to compute because $0 or $-n could appear yes, the general problem is probably impossible > in semantic actions at any level in the parse tree. Bison would have to > check every possible expansion of every possible RHS symbol. That > implementation seems like more work than I want to do. ... but how hard would it be to check just one level ? > > What if we simply add an option to turn this warning off globally? I think that for my development, I will just do a dummy reference to $3 Henrik _______________________________________________ help-bison@gnu.org http://lists.gnu.org/mailman/listinfo/help-bison